直升机主减速器弱监督祛工况故障诊断方法、装置及设备
By combining time-frequency analysis and neural networks, single-component and multi-component time-frequency feature spectra are generated, and weakly supervised meshing frequency ridge extraction is performed. This solves the problems of noise sensitivity and poor adaptability to nonlinear systems in order tracking methods without tachometers, and achieves higher robustness, accuracy, and better adaptability in fault diagnosis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- AECC HUNAN AVIATION POWERPLANT RES INST
- Filing Date
- 2024-04-25
- Publication Date
- 2026-07-17
AI Technical Summary
Existing tachometer-less order tracking methods are sensitive to noise, prone to errors, and have poor adaptability to nonlinear and non-stationary systems, lacking versatility and increasing the complexity of practical engineering applications.
A method combining time-frequency analysis and neural networks is adopted. Single-component and multi-component time-frequency feature spectra are generated through synchronous compression transformation and multiple synchronous compression transformations. The vibration transmission path is fitted by matrix, and the generated single-component reference signal feature spectrum is used. A fault diagnosis method based on planetary network is adopted. The weakly supervised meshing frequency ridge line is extracted by neural network, the instantaneous frequency trend line is synthesized and the signal angle domain is resampled to achieve spectrum analysis.
It improves robustness to noise, reduces errors, enhances adaptability to nonlinear and non-stationary systems, improves the accuracy and reliability of diagnostic results, has good versatility, and reduces the complexity of practical engineering applications.
Smart Images
Figure CN118673299B_ABST